Midjourney is a great AI tool to brainstorm design ideas and today I want to show you some things ive learned recently, like how to correctly structure a text prompt to get the best results, also how to use image references in the correct way to influence your image, and finally how to use additional tweaking and prompt formatting to elevate your image to a whole new level.
In MidJourney, crafting the perfect prompt involves a clear understanding of three key elements: Medium, Subject, and Environment. By simplifying our prompts to focus on these aspects, we can guide the AI more effectively, ensuring that our architectural visualizations are not only beautiful but also precise. This approach allows us to maintain coherence throughout our images and maximize MidJourney's powerful capabilities.
Medium: This refers to the visual style or technique you want the AI to simulate in the image. For architectural visuals, the medium can dramatically affect the perception of the design. For example:
When creating architectural images that closely mimic photographs, specifying the 'Medium' in your prompt is crucial. The choice of words can guide MidJourney to simulate different styles of photography, each capable of highlighting unique aspects of architecture. Here’s how to describe various photographic mediums:
When I use this prompt, and change the medium, the whole image changes. If i use the words rendering, its going to look very different than using the word 35 mm.
If you are going for a realistic camera feel, you might want to use cameras that architects typically use to take pictures. Some of my favourite words I like to use are, Photograph of (insert subject) shot on a large format camera. Or shot on a nikon D700 camera, etc.
If you want a realistic camera, but with a more vintage feel, you can search what film stocks are used for architecture photography and also have interesting results. Some of my favourites are, Kodak Ektar 100 film still, Kodak Hawkeye Traffic Surveillance Film 400 film still, Kodak Portra 400 film still among others.
Other words you might want to use as a medium are, landscape photography, urban photography, aerial photography, architectural detail photography.
Now lets talk about subject;
This is the central focus of your image. In the context of architecture, this includes not only the building itself but also specific elements that define its character. For example:
When crafting prompts for architectural images, the 'Subject' component is key to directing MidJourney’s focus on particular aspects of a building or structure. This includes the overall architectural design, specific elements, and materials.
And when talking about overall building design, word precision is important. Here I always recommend to grab your favourite architecture book and read the descriptions to understand and find specific words that can work for you. You can use the words, modern, victorian or brutalist, which can work well. But you can also be more specific and have words like parametric, organic, deconstructivis, sculptural, pixelated, monolithic, modular, lattice, etc.
I recommend you also specific structural elements, like cantilevered balconies, glass facade, steel beams. Facade treatments, like brickwork, cladding, corten steel. Material specificty like polished concrete, wood, corrugated metal. And distinct architectural features like rooftop garden, floating staircase, intricate masonry.
Now lets talk about, Environment: Which is The setting in which the architecture is placed, which frames the subject and provides context. This can alter the mood and narrative of the image. Key considerations include:
The 'Environment' in an architectural prompt sets the stage, providing context that influences the emotional and aesthetic impact of the image. Here's how to effectively use environmental descriptions in your prompts:
Urban vs. Rural Settings:
2. Specific Landscapes:
3. Time of Day:
4. Weather Conditions:
5. Cultural or Historical Context:
6. Seasonal Influence:
By carefully choosing and describing the environment in your prompts, you can direct the visual narrative, enhancing the architectural subject with a richly contextual backdrop that complements or contrasts its design
Now lets take a look at some example prompts.
If I wanted to design a university in france, I would structure my prompt in the following way:
“exterior photograph of a university building in a serene suburban setting in France shot on a large format camera. The building features a unique architectural style with a recessed ground floor and cantilevered glass upper floors, creating a floating effect. The roof of the building serves as a multi-sport court enclosed by a tall metal fence. The building structure is entirely of exposed concrete, with each floor slab exposing concrete beams beneath it. There are a group of students entering the building in a white and blue uniform. The scene captures a tranquil and airy atmosphere, enhanced by soft daylight, soft shadows and a natural color palette”
Here you can identify the medium, the subject, and the environment.
A poor example of this prompt would look something like this:
A building of a university in france, the building is modern and the day is sunny, in the style of Rem Koolhaas.
Now I think that to be good at prompting, we have to learn how to prompt and use trigger words, and not always use reference architects. We know using the words frank ghery, mies van der rohe, zaha hadid, will generate a specific image of a building. But if we want to generate some quote unquote original and different results, lets try to be structured and descriptive with our base prompt.
Now that we have our image ready, lets try to be more specific and use some image references.
Inside of midjourney we can use image references to influence our base image, all we have to do is at the end of the prompt, include the words —sref and the link to the image address and thats it. So for example, I will grab some photographs of my iphone and use them as references. I will upload them to discord, and copy the image address and then paste it.
Now If I use the same prompt and just add the image reference, this is the result I will have. If you see that the influence of the image is too strong, you and use the words —sw and specificy a number, by default its in 100. So I can lower it to 50, or if I want more influence I can leave it at 200.
Now you can also use an image reference to influence your color palette. This is a really cool trick and its really easy. I recommend you do this after you have your almost ready image. So we have worked on this school in various tries and lets say this is the final point of view and image we have. Now I am going to go to pinterest, and type in color palette. But if you have a specifc one you can use that one too.
We are going to click on remix our image, apply the style reference, plus the link, and in the influence of the image we are going to put 50. I usually leave it at a low influence, here are different images I exported where I just changed the color palette. This is one of the coolest features if you ask me.
Now all of these prompts can be adjusted by using the settings, where you can adjust the image size and ratio, select the mode and version, choose the aesthetics and also change speed modes. Lets walk through each one of these very quick.
First the image ratio, at first you would just think its the image ratio. But choosing this before hand will also influence the output of your image. If your medium is a photography medium, and you choose a 1.1 square ratio, the chances of having a more vintage look are higher than if you choose a vintage film but select a landscape ratio.
Also, increasing the level of stylization makes the images more like previous versions of midjourney.
You can play with all of these settings, but I usually leave them as it is, use a landscape ratio, use RAW mode, use the latest version, and thats it.
When you have your image ready you can also play with some settings like zooming out, panning, and remixing, but all of that is for another video.